What stable overclocking setups are you using on the R7 3700X and R7 3800X?
What stable overclocking setups are you using on the R7 3700X and R7 3800X?
If I remember correctly, CaseKing also opts for less demanding stress test settings by employing a Prime95 version before the intensive AVX operations and using substantial FFT data sets. This approach provides a more authentic 'real-world' scenario, better representing actual heavy usage than the smaller FFTs with many AVX instructions. CaseKing also uses (award-winning overclocker) der8auer for their binning process, which you should know about.
I have a 3700X and use it with an Asus ROG Strix X570-E Gaming board. I can run my chip at 4.3ghz @ 1.35v with an LLC of 3 or 4. It works well and stays stable, especially at LLC 3, though I can go down to 4 if needed. This setup is air-cooled.
I think my chip could likely reach 4.35ghz+ by increasing the voltage slightly. But for now I’m satisfied with 4.3ghz/1.35v.
Additionally, many system monitoring tools display "1.48v" or "1.5v" only because they are accounting for the maximum voltage being applied to any one core. Thus, during single-threaded tasks, "1.5v" is limited to a single core and isn't as risky as some people think. It's simply typical behavior when you run your chip in 'auto' mode.
Read back some of these points, as the Stilt comments clarify what’s going on. When voltage reaches between 1.45 and 1.5 volts it falls into periods with minimal current flow, even though clock speeds are high and only a light workload is handled. This combination of high voltage and low current density isn’t dangerous; it’s the simultaneous presence of both that can cause issues. The system will automatically adjust the voltage and clock speeds back to what manual overclockers use when similar conditions arise. A manual overclock at low voltage won’t allow you to reach the processor’s maximum clock speeds with brief, fluctuating loads—exactly the kind of scenario games present.
If the safe maximum voltage is 3800x with a 1.3V core voltage, manual overclocking isn't necessary. However, with a 1.4V limit, the 3800x could significantly outperform most 9900ks CPUs using a gold sample. For 4.4GHz+ overclocks with 1900 and 3800 RAM CL15, you can achieve a time sky CPU score of 11700, matching the average score of 9900ks CPUs.
Whoops... yes, I actually missed mentioning the low current / light workload part as well. Thanks. And yeah, I’ll have to go back and re-read the whole thread and article. But yeah, many people are surprised when they see voltage readings near 1.5... but in the end, it’s nothing to worry about (especially with Ryzen 3000) and I’ve seen several well-respected community members comment on this topic and they all agree. Thanks for the correction.
I'm not really sure I should care. Even my 3700X scores so close to a 9900ks it's indescernible in games. And it already beats one in content creation or productivity (except a few Adobe apps).
And what's more, play it in a realistic setup...meaning NOT on a $1200 GPU at absurdly low 1080P resolution...it will score the same. It's the GPU that makes the difference; even with a 5700XT playing at 1440P my 3700X is just loafing along at 30-40% MAX utilization in Ghost Recon BP.
There's just no compelling reason to enter into the grief and bother of balancing the knife edge of an all-core, low voltage overclock.
3700x doesn’t match a 9900ks in games. Achieving an average of 9900ks requires a 3800x on time. The 9900ks offer significantly more performance than the 3800x. You can obtain 3600 RAM with adjusted timings and achieve a score above 12k for a time spy cpu. The 3700x is generally slower overall, making it unable to compete with the 3800x in most cases due to binning effects. Many 3700x models struggle with higher IF frequencies. Reviews claiming a 3700x overclocked as a 3800x aren’t accurate.
On the other hand, you can push a 3700x to its limits and a gold sample reach 4.4GHz across all cores. Some models support IF up to 1900. With 3733 RAM and tight timings, high scores are possible. Here’s a link: https://www.3dmark.com/spy/9271166
This 3700x has a CPU score of 12,187, ranking number 7 globally—this isn’t a daily average. Some enthusiasts run bench tests using high voltage. Both the 3800x and 3700x rely heavily on RAM timings, not just clock speeds. The 9900ks can also be overclocked similarly.
I’m puzzled about how the top performer achieved a 14k time spy cpu score with a 3700x and 3200 RAM.
I'm sorry, but it's hard to see exactly what a 9900ks could achieve given my 3700x running smoothly at 30-40% load while streaming GR-BP in 1440p. Even with a 2080ti, things would only get worse at 4k. When you mention "can't match..." it's mostly theoretical.
And yes, everything is VH to ultra—love the visuals.
Overclocking a 3800x to narrow the performance gap caused by an expensive processor is also a challenge. I just want something that actually helps me out of this situation. I'm not into the kind of competition where I'm just trying to prove who can go further than others, like a kid at scout camp.